Questions and Answers on Auto-Pilot

In today’s day and age, we have the privilege of getting answers in real time to the questions we ask. And many of the answers we seek are now powered by data-driven insights we train our AI models to produce. A recent episode of Inside Analysis featured industry experts’ sharing their thoughts on the field of AI’s development.? Host Eric Kavanagh interviewed Alex Kangoun, PMP of Athena Technology Solutions and Ryan Welsh of Qlik about the ways in which their emerging technologies can address organizations’ needs.?

Pooling Resources and Navigating Structure?

Kangoun’s presentation at the start of the webinar highlighted a crucial aspect of data management: "For data to be anonymized and encrypted, it takes time and processing power,” he said.? Anonymizing and encrypting data are essential steps in protecting privacy and securing sensitive information. However, these processes are resource-intensive. The computational power required to ensure data privacy often slows down data processing, posing a significant challenge for businesses aiming to handle large volumes of data efficiently.

Moreover, the size of the data is not the only issue organizations have to consider. As Welsh said,?

"Unstructured content makes up 80 percent of the world’s data, and users spend about 20 percent of their time looking for answers in unstructured content.” Unstructured data, which includes everything from emails and social media posts to videos and audio files, can play a large role in providing necessary context surrounding any project. However, not many businesses are currently able to work with unstructured data as smoothly as they can with structured data. Qlik Answers, as explained by Welsh, is designed to bridge this gap by providing users with quick and accurate responses, regardless of the data's structure.

Such a seamless interaction between structured and unstructured data is crucial. As Welsh put it, "Users don’t care where the answers to their questions come from. They just want the answer." Users are not particularly forgiving of a difference in data types’ causing a lag in their process flow. They require solutions that can handle diverse data types seamlessly, improving the efficiency and effectiveness of their data-driven decision-making.

Integrating Large Language Models (LLMs)

LLMs are quickly becoming a solution that users turn to not only for business but in their daily lives. This visibility puts significant pressure on LLMs to perform perfectly and Kavanagh provided a necessary reality check, saying, "One of the biggest misconceptions about LLMs is that they have to be 100 percent correct to have any value, but the old way wasn’t 100 percent correct and curation will always be part of the process."?

The notion that LLMs need to achieve perfect accuracy overlooks the reality that even traditional methods of data analysis were not flawless. The value of LLMs lies in their ability to process vast amounts of data quickly and generate useful insights, even if those insights occasionally require human curation.

Kavanagh also highlighted the importance of prompt engineering while working with LLMs: "The smallest changes in your prompt can have significant effects on the results that you get,” he said, emphasizing the nuanced nature of working with LLMs. A well-crafted prompt can significantly enhance the quality of the output, making prompt engineering a critical skill for anyone working with these models.

AI in the Workforce

Widespread experimentation with AI models in a casual fashion could lessen fears that AI will completely replace humans in the workforce. By interacting with chatbots and seeing firsthand how prompts influence outcomes, users can witness the sustained value of human input.?

As Welsh suggested, "It’s not so much that AI will replace humans in the workforce, but people who use AI in their workflows will replace people who are not comfortable using AI in their workflows." This perspective seeks to replace fear of the unknown with the motivation to experiment and get involved. In the short term, we may begin to witness that a competitive advantage from those who embrace AI as compared to those who do not.

Welsh also touched on the need for innovative architectures to develop advanced reasoning systems: "To have a really good reasoning system, you might have to create a new architecture,” he said. Current systems, while powerful, still lag behind human cognitive abilities, particularly in their ability to switch contexts quickly. As Kavanagh notes, "Humans can jump tracks in their mind, and these systems aren’t quite there yet."

Conclusion

At the end of the webinar, Welsh encouraged organizations to empower their employees to experiment with AI systems. Not only does this empowerment lead to a more open culture, but it could actually spark new and profitable business. "Enable your people to play with these systems because they will uncover use cases that you’ve never thought of,” he urged.?

In other words, fostering innovation at every stage of an organization’s pipeline might involve an investment in time, but is likely to come out net positive in terms of revenue and innovation. As we learn to interact with AI as part of our daily workflows, we will prove that AI is not only reserved in a high-brow capacity for top-level executives. In fact, there are tasks at every level of a business that could benefit from AI interaction.?